Method and apparatus for segmenting and classifying unconstrained handwritten characters

Abstract

A character recognition system including a method and apparatus for pre-processing handwritten characters prior to segmentation and classification is disclosed. The method includes the steps of forming one or more sub-regions each corresponding to a single character, accumulating the subregions to form an invalid region, and segmenting each character in the invalid region to locate said corresponding characters. The method is included as part of an apparatus for recognizing handwritten text.

I claim: 
     
       1. A method for recognizing an object contained within an image, the method comprising the steps of; generating an electronic bitmap, comprising a plurality of pixels, representative of said image;   identifying a group of said pixels, each having a grey-level value associated therewith, associated with said object;   enclosing a region of said bitmap including said group of pixels within an invalid region, wherein said enclosing step comprises the step of determining an optimum portion of said bitmap needed to segment said object, wherein said determining step comprises the step of ascertaining whether said object overlaps a second object in said image;   processing said invalid region to thereby segment said object; and   thereafter classifying said segmented object.   
     
     
       2. The method of claim 1 wherein said determining step further comprises the steps of: determining an optimum second portion of said bitmap which encloses said second object; and   accumulating said second portion of said bitmap into said invalid region.   
     
     
       3. The method of claim 1 wherein said invalid region comprises a list of one or more first enclosing zones, wherein each of said one or more first enclosing zones encloses a portion of said object. 
     
     
       4. The method of claim 1 wherein said determining step comprises determining the smallest portion of said bitmap needed to segment said object. 
     
     
       5. The method of claim 1 wherein said invalid region comprises a list of one or more first enclosing zones, wherein each of said one or more first enclosing zones encloses a portion of said object, and wherein said accumulating step comprises the step of concatenating said list of one or more first enclosing zones with a list of one or more second enclosing zones associated with said second object. 
     
     
       6. A method for classifying a character in a bitmap, said character comprising indicia of one or more pen strokes, said method comprising the steps of: forming an invalid region about said character, wherein said step of forming said invalid region comprises the step of determining a smallest portion of said bitmap needed to enclose said character, wherein said determining step comprises the step of determining whether said character overlaps a previously classified character in said bitmap;   segmenting said character using said invalid region; and   thereafter classifying said character.   
     
     
       7. The method of claim 6 wherein said determining step further comprises the steps of: determining a smallest portion of said bitmap needed to enclose said overlapped character; and   concatenating said smallest portion of said bitmap needed to enclose said character into said smallest portion of said bitmap needed to enclose said overlapped character to form an updated invalid region.   
     
     
       8. The method of claim 7 wherein said smallest portion of said bitmap needed to enclose said overlapped character comprises a list of one or more second enclosing zones, wherein each of said one or more second enclosing zones encloses one or more pen strokes associated with said overlapped character. 
     
     
       9. The method of claim 8 wherein said forming step comprises the step of accumulating said list of one or more second enclosing zones into said list of one or more first enclosing zones. 
     
     
       10. The method of claim 6 wherein said forming step comprises the step of creating a list of one or more first enclosing zones, wherein each of said one or more first enclosing zones encloses one of said one or more pen strokes of said character. 
     
     
       11. Apparatus for recognizing handwritten text on a background, said handwritten text including one or more characters each including one or more pen strokes, said apparatus comprising: means for forming a list of one or more first regions, each of said one or more first regions corresponding to one of said one or more pen strokes;   means for segmenting said list of one or more first regions to extract said one or more characters of said handwritten text from said background;   means for classifying one or more characters to recognize said handwritten text;   means for forming a list of one or more second regions, each of said one or more second regions corresponding to one of said one or more pen strokes of a previously segmented or classified character which is overlapped by said one or more characters; and   means for accumulating said list of one or more second regions into said list of one or more first regions to update said list of one or more first regions.